基于PCA-GA-SVR的鲜食葡萄运输过程品质建模

doi:10.19788/j.issn.2096-6369.220111

摘要/Abstract

摘要：

由于生鲜果蔬生产的地域性和季节性，生鲜果蔬在采摘后需要经过运输、贮藏等物流过程才可以到达消费者的手中，在此过程外界因素的影响会造成生鲜果蔬发生一系列生理变化，进而影响其口感。本研究以鲜食葡萄为研究对象，拟探索鲜食葡萄在物流运输过程中的理化指标与感官品质关系建模，基于对实际运输过程的监测在实验室开展了鲜食葡萄运输模拟实验和感官实验，获取了鲜食葡萄运输过程中理化指标和感官品质数据，并构建了不同运输模式下的鲜食葡萄品质数据集；构建了基于改进支持向量回归（Support Vector Regression Algorithm，SVR）的鲜食葡萄理化指标与感官品质建模方法，首先利用主成分分析法（Principal Component Analysis， PCA）对理化指标进行降维，再利用遗传算法（genetic algorithm， GA）优化SVR模型参数提升模型的拟合效果。利用常温运输、保冷运输和冷链运输三种运输模式以及混合数据集测试，结果表明改进的PCA-GA-SVR模型预测的准确性和精度均有显著提高，MAE、MSE、RMSE均小于0.5，R2均大于0.96，PCA-GA-SVR模型能够较好地反映鲜食葡萄理化指标与感官品质之间的非线性映射关系。同时，研究表明理化指标数据与感官品质之间的关系受到运输模式的影响较小，本研究提出的感官品质评估模型可以较好地应用在任何运输方式上，辅助物流过程中生鲜农产品的品质控制与管理。

关键词: 多支持向量回归, 理化指标, 感官品质, 遗传算法, 鲜食葡萄, 生鲜农产品, 食品安全

Abstract:

The regionality and seasonality of fresh fruits and vegetables mean that this produce is subjected to logistics and transportation processes after harvesting and before reaching the hands of consumers. During these processes, several external factors cause physiological changes in fresh fruits and vegetables, which will affect their taste. This paper considers the case of table grapes, and conducts simulation experiments and sensory experiments in the laboratory based on monitoring of the actual transportation process. After obtaining a suitable dataset, a model of the relationship between the physicochemical indicators and the sensory quality of the grapes is constructed based on an improved support vector regression (SVR) algorithm. Principal component analysis (PCA) is used to reduce the dimensionality of the physical and chemical indicators, and a genetic algorithm (GA) is used to optimize the SVR model parameters. An examination of three transportation modes (normal-temperature transportation, cold-storage transportation, and cold-chain transportation) and mixed dataset testing shows that the improved PCA-GA-SVR model offers significantly improved accuracy and precision, reflecting the nonlinear mapping relationship between the physicochemical indexes and sensory quality of table grapes. The mean absolute error, mean squared error, and root mean squared error achieved by the PCA-GA-SVR model are all less than 0.5, and the R2 values are all greater than 0.96. This study shows that the relationship between physical and chemical index data and sensory quality is not significantly affected by the transportation mode. The sensory quality evaluation model proposed in this study can be applied to any transportation mode to assist the quality of fresh agricultural products in the control and management of the logistics process.

Key words: Multiple Support Vector Regression, physical and chemical indicators, sensory quality, Genetic Algorithm, table grapes, fresh agricultural products, food safety

中图分类号:

TU205

贺苗, 李鑫, 朱志强, 冯建英. 基于PCA-GA-SVR的鲜食葡萄运输过程品质建模[J]. 农业大数据学报, 2022, 4(1): 98-108.

Miao He, Xin Li, Zhiqiang Zhu, Jianying Feng. Modeling Table Grapes: Physicochemical Indexes and Sensory Quality Based on PCA-GA-SVR during Transportation Process[J]. Journal of Agricultural Big Data, 2022, 4(1): 98-108.

图/表 13

图1

表1

图2

表2

表3

表4a

常温运输PCA-GA-SVR组合模型评价结果"

	MAE	MSE	RMSE	$R 2$
SVR	1.139	1.749	1.323	0.838
PCA-SVR	0.948	1.307	1.143	0.879
PCA-GA-SVR	0.461	0.186	0.431	0.965

表4a

图3

表4b

保冷运输PCA-GA-SVR组合模型评价结果"

	MAE	MSE	RMSE	$R 2$
SVR	1.273	2.062	1.436	0.787
PCA-SVR	0.773	1.031	1.016	0.867
PCA-GA-SVR	0.391	0.184	0.429	0.978

表4b

表4c

冷链运输PCA-GA-SVR组合模型评价结果"

	MAE	MSE	RMSE	$R 2$
SVR	1.134	2.031	1.415	0.787
PCA-SVR	0.775	0.986	0.993	0.827
PCA-GA-SVR	0.330	0.157	0.396	0.981

表4c

表4d

混合数据集PCA-GA-SVR组合模型评价结果"

	MAE	MSE	RMSE	$R 2$
SVR	1.141	1.069	1.034	0.838
PCA-SVR	0.824	0.649	0.806	0.914
PCA-GA-SVR	0.357	0.124	0.352	0.983

表4d

图3

参考文献 30

1	Akaberi M, Hosseinzadeh H. Grapes (Vitis vinifera) as a Potential Candidate for the Therapy of the Metabolic Syndrome[J]. Phytotherapy Research, 2016,30(4):540-556.
2	原变鱼. 鲜食葡萄可持续供应链风险评价的神经网络模型优化及系统[D].中国农业大学，2021.
	Yuan B Y. Neural network model optimization and system for risk assessment of sustainable supply chain of table grapes [D]. China Agricultural University, 2021.
3	王康飞,王桂英,王德铮.不同保鲜方式对葡萄保鲜效果影响的对比研究[J].包装工程,2020,41(15):19-24.DOI:10.19554/j.cnki.1001-3563.2020.15.004 . doi: 10.19554/j.cnki.1001-3563.2020. 15.004
	Wang K F, Wang G Y, Wang D Z. A comparative study on the effect of different preservation methods on the effect of grape preservation [J]. Packaging Engineering, 2020, 41(15): 19-24. DOI: 10.19554/j.cnki.1001-3563.2020. 15.004 . doi: 10.19554/j.cnki.1001-3563.2020. 15.004
4	HONG H S P S J. H.. Effect of Familiarity on a Cross-Cultural Acceptance of a Sweet Ethnic Food: A Case Study with Korean Traditional Cookie[J]. Journal of Sensory Studies, 2014,29:110-125.
5	李鑫.基于支持向量回归建模的鲜食葡萄运输感官品质评价与预测[D].中国农业大学，2021.
	Li X. Evaluation and prediction of sensory quality of table grape transportation based on support vector regression modeling [D]. China Agricultural University, 2021.
6	Rajamaki T, Alakomi H, Ritvanen T, et al. Application of an electronic nose for quality assessment of modified atmosphere packaged poultry meat[J]. Food Control, 2006, 17(1): 5-13.
7	徐赛,陆华忠,周志艳,吕恩利,杨径.基于电子鼻的果园荔枝成熟阶段监测[J].农业工程学报,2015,31(18):240-246.
	Xu S, Lu H Z, Zhou Z Y, et al. Monitoring of orchard litchi maturity stage based on electronic nose [J]. Chinese Journal of Agricultural Engineering, 2015, 31(18): 240-246.
8	HEMPEL A, O'SULLIVAN M.G,et al. Use of optical oxygen sensors to monitor residual oxygen in pre- and post-pasteurised bottled beer and its effect on sensory attributes and product acceptability during simulated commercial storage.[J]. LWT-Food Science & Technology,2013(1). DOI:http://dx.doi.org/10.1016/j.lwt.2012.05.026 . doi: 10.1016/j.lwt.2012.05.026
9	Dong C, Zhu H, Zhao J, et al. Sensory quality evaluation for appearance of needle-shaped green tea based on computer vision and nonlinear tools[J]. Journal of Zhejiang University-SCIENCE B, 2017, 18(6): 544-548.
10	董春旺,朱宏凯,周小芬,袁海波,赵杰文,陈全胜.基于机器视觉和工艺参数的针芽形绿茶外形品质评价[J].农业机械学报,2017,48(09):38-45.
	Dong C W, Zhu H K, Zhou X F, et al. Evaluation of the shape and quality of needle bud-shaped green tea based on machine vision and process parameters [J]. Journal of Agricultural Machinery, 2017, 48(09): 38-45.
11	唐雪平. 基于化学分析与机器学习的铁观音茶叶品质评价体系[D].华侨大学,2020.DOI:10.27155/d.cnki.ghqiu.2020.000700 . doi: 10.27155/d.cnki.ghqiu.2020.000700
	Tang X P. Quality evaluation system of Tieguanyin tea based on chemical analysis and machine learning [D]. Huaqiao University, 2020. DOI: 10.27155/d.cnki.ghqiu.2020.000700 . doi: 10.27155/d.cnki.ghqiu.2020.000700
12	罗娜. 数据挖掘中的新方法——支持向量机[J]. 软件导刊, 2008(10):30-31.
	Luo N. A New Method in Data Mining——Support Vector Machine [J]. Software Guide, 2008(10):30-31.
13	Hsu C W, Lin C J. A simple decomposition method for support vector machines[J]. Machine Learning, 2002, 46(1): 291-314.
14	蒋华伟,陈斯,杨震.基于IGWO-SVR短期储藏小麦品质预测模型研究[J].中国粮油学报,2021,36(08):79-87.
	Jiang H W, Yang S, Yang Z. Research on the quality prediction model of short-term storage wheat based on IGWO-SVR [J]. Chinese Journal of Cereals and Oils, 2021, 36(08): 79-87.
15	Yao X, Cai F, Zhu P, et al. Non-invasive and rapid pH monitoring for meat quality assessment using a low-cost portable hyperspectral scanner[J]. Meat science, 2019, 152: 73-80.
16	Wang Y, Li T, Li L, et al. Micro-NIR spectrometer for quality assessment of tea: Comparison of local and global models[J]. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 2020,237:118403.
17	Tang X L, Li Y, Li P, et al. Model predictive control based on SVR optimized by multi-agent particle swarm optimization algorithm[J]. Control and Decision, 2014, 29(4): 593-598.
18	Zhang HQ, Chen CX, Lu YFet al.Application of IWO-PSO-SVR algorithm in methane[J].China Environmental Science,2020(40):1453-1459.
19	孙俊,莫云南,戴春霞,陈勇,杨宁,唐游.基于介电特性与IRIV-GWO-SVR算法的番茄叶片含水率检测[J].农业工程学报,2018,34(14):188-195.
	Sun J, Mo Y N, Dai C X, et al. The tomato leaves based on the characteristics of the dielectric and Iriv-Gwo-SVR algorithm test [J]. 14: 188-195.
20	肖建,于龙,白裔峰.支持向量回归中核函数和超参数选择方法综述[J].西南交通大学学报,2008(03):297-303.
	Xiao J, Yu L, Bai Y F. A review of kernel function and hyperparameter selection methods in support vector regression [J]. Journal of Southwest Jiaotong University, 2008(03):297-303.
21	崔东文,金波.基于改进的回归支持向量机模型及其在年径流预测中的应用[J].水力发电学报,2015,34(02):7-14.
	Cui D W, Jin B. An improved regression support vector machine model and its application in annual runoff prediction [J]. Journal of Hydroelectric Power Generation, 2015, 34(02): 7-14.
22	Liu F, He Y, Wang L. Comparison of calibrations for the determination of soluble solids content and pH of rice vinegars using visible and short-wave near infrared spectroscopy[J]. Analytica Chimica Acta, 2008,610(2):196-204.
23	Lin H, Chen Q, Zhao J, et al. Determination of free amino acid content in Radix Pseudostellariae using near infrared (NIR) spectroscopy and different multivariate calibrations[J]. Journal of pharmaceutical and biomedical analysis, 2009, 50(5): 803-808.
24	王雪冰, 郭庆胜, 王勇, 等. 地图图片的特征提取与自动识别方法[J]. 测绘与空间地理信息, 2019,42(09):28-32.
	Wang X B, Guo Q S, Wang Y, et al. Feature extraction and automatic identification method of map images [J]. Surveying and Mapping and Spatial Geographic Information, 2019,42(09):28-32.
25	王敬. 文本分类中SVM核函数的探讨[D].兰州大学,2021.DOI:10.27204/d.cnki.glzhu.2021.002135 . doi: 10.27204/d.cnki.glzhu.2021.002135
	Wang J. Discussion on SVM Kernel Function in Text Classification [D]. Lanzhou University, 2021. DOI: 10.27204/d.cnki.glzhu.2021.002135 . doi: 10.27204/d.cnki.glzhu.2021.002135
26	Wu D, Jiang Z, Xie X, et al. LSTM learning with Bayesian and Gaussian processing for anomaly detection in industrial IoT[J]. IEEE Transactions on Industrial Informatics, 2019, 16(8): 5244-5253.
27	Li Y, Jia X, Wang R, et al. A new oversampling method and improved radial basis function classifier for customer consumption behavior prediction[J]. Expert Systems with Applications, 2022: 116982.
28	Gao W, Han J. Prediction of Destroyed Floor Depth Based on Principal Component Analysis (PCA)-Genetic Algorithm (GA)-Support Vector Regression (SVR)[J]. Geotechnical and Geological Engineering, 2020,38(4):3481-3491.
29	张凌栩,韩锐,李文明,史银雪,刘驰.大数据深度学习系统研究进展与典型农业应用[J].农业大数据学报,2019,1(02):88-104.DOI:10.19788/j.issn.2096-6369.190208 . doi: 10.19788 /j.issn.2096-6369.190208
	Zhang L Y, Han R, Ling W M, et al. Research progress and typical agricultural applications of big data deep learning systems [J]. Journal of Agricultural Big Data, 2019, 1(02): 88-104. DOI: 10.19788 /j.issn.2096-6369.190208 . doi: 10.19788 /j.issn.2096-6369.190208
30	周平，李瑞峰, 郭东伟, 等. 高炉炼铁过程多元铁水质量指标多输出支持向量回归建模[J].控制理论与应用, 2016,33(06):727-734.
	Zhou P, Li R F, Guo D W, et al. Multi-output Support Vector Regression Modeling of Multiple Hot Metal Quality Indicators in Blast Furnace Ironmaking Process [J]. Control Theory and Application, 2016,33(06):727-734.

感官属性（满分100分）	评价标准	打分范围
外观（满分20分）	果穗松散、落粒，果梗干枯，果形差，果粒萎蔫、流汁、无果粉	1-5
	果穗较紧实，果梗稍褐变，果形较好，果色暗淡、果粉部分脱落	6-15
	果穗紧密适度、果梗鲜嫩、果形端正，果色均匀、布满果粉	16-20
香气（满分20分）	无品种特定香气或香气不明显	1-5
	有品种特定香气，但是香气较淡	6-15
	具有浓郁的葡萄品种特定香气	16-20
果皮和果肉质地（满分30分）	果皮粗糙，果肉黏滑，果皮、果肉呈袋状分离	1-10
	果皮韧性大于脆性，果肉变软	11-20
	果皮膨压大，食用时易碎裂，果肉紧厚而不粗糙，多汁	21-30
果粒风味（满分30分）	风味不协调，酸度过大，涩味重，有不良气味	1-10
	风味较好，酸甜味较淡，无不良气味	11-20
	风味极佳，糖酸比例协调，具有葡萄特有的芳香	21-30

指标	失重率	可滴定酸含量	可溶性固形物含量	抗坏血酸含量	果梗拉力	果梗叶绿素含量	硬度	弹性	凝聚性	咀嚼性	回复性	感官评价
失重率	1	0.49	0.36	0.50	-0.91	-0.96	-0.91	-0.88	-0.91	-0.93	-0.84	-0.87
可滴定酸含量	0.49	1	0.57	0.83	-0.41	-0.34	-0.43	-0.41	-0.47	-0.42	0.36	-0.33
可溶性固形物含量	0.36	0.57	1	0.59	0.24	0.23	0.18	-0.23	-0.27	-0.19	-0.14	-0.44
抗坏血酸含量	0.50	0.83	0.59	1	-0.40	-0.35	-0.40	-0.42	-0.46	-0.40	-0.34	-0.33
果梗拉力	-0.90	-0.40	0.24	-0.41	1	0.91	0.89	0.85	0.87	0.91	0.85	0.82
果梗叶绿素含量	-0.96	-0.34	0.24	-0.35	0.92	1	0.92	0.88	0.89	0.93	0.86	0.88
硬度	-0.91	-0.43	0.18	-0.40	0.89	0.91	1	0.85	0.88	0.98	0.85	0.88
弹性	-0.88	-0.41	-0.23	-0.42	0.85	0.87	0.8	1	0.83	0.89	0.76	0.79
凝聚性	-0.91	-0.46	-0.26	-0.46	0.87	0.89	0.88	0.83	1	0.93	0.83	0.83
咀嚼性	-0.93	-0.42	-0.19	-0.40	0.91	0.93	0.98	0.89	0.94	1	0.86	0.88
回复性	-0.84	-0.36	-0.14	-0.34	0.85	0.86	0.85	0.76	0.83	0.86	1	0.80
感官评价	-0.86	-0.33	-0.44	-0.32	0.82	0.88	0.87	0.79	0.83	0.88	0.81	1

主成分	特征差	特征差贡献率%	累计特征差贡献率%
Z1	6.7604	61.379	61.379
Z2	2.7630	25.087	86.466
Z3	0.4428	4.020	90.486
Z4	0.2805	2.547	93.033
Z5	0.1974	1.792	94.825
Z6	0.1556	1.412	96.237
Z7	0.1325	1.203	97.440
Z8	0.1137	1.032	98.472
Z9	0.0976	0.886	99.358
Z10	0.0691	0.628	99.986
Z11	0.0015	0.014	100.000

[1]	李佳利, 杨涵, 钱建平. 新冠肺炎疫情下中国农产品冷链物流可信追溯研究[J]. 农业大数据学报, 2022, 4(1): 14-24.
[2]	张哲, 杨信廷, 于合龙, 李珊珊, 孙传恒. 基于区块链技术的生鲜农产品追溯系统研究进展[J]. 农业大数据学报, 2022, 4(1): 25-34.
[3]	张新, 彭祥贞, 李悦, 王小艺, 赵峙尧, 许继平. 基于可信区块链和可信标识的粮油食品全供应链信息追溯模型[J]. 农业大数据学报, 2022, 4(1): 35-48.
[4]	周大森, 陈静, 张鑫, 李继兰, 李志成, 郑晓冬, 公维敏, 宋烨. 果蔬冷链追溯体系研究现状与应用建议[J]. 农业大数据学报, 2022, 4(1): 49-54.
[5]	张茜, 田乙慧, 肖文, 鲁燕. 大数据在农产品冷链物流中的应用[J]. 农业大数据学报, 2022, 4(1): 55-61.
[6]	刘佳, 夏晓蕾, 王丽娟, 王姝, 吕雪峰, 路长发, 孙利, 李素彩. 进口冷链食品追溯管理平台及创新应用[J]. 农业大数据学报, 2022, 4(1): 69-76.
[7]	陈晨. 新冠疫情下基于区块链技术的进口冷链食品追溯体系构建研究[J]. 农业大数据学报, 2022, 4(1): 77-81.
[8]	陈谦, 杨涵, 王宝刚, 李文生, 钱建平. 基于GRU神经网络模型的冷链运输温度时序预测[J]. 农业大数据学报, 2022, 4(1): 82-88.
[9]	周蓉蓉, 陈栋, 刘思远. 基于K均值聚类算法的生鲜运输路径优化模型[J]. 农业大数据学报, 2022, 4(1): 89-97.
[10]	陈志军, 刘艳, 钱永忠. 基于风险熵的农产品安全定量评价研究[J]. 农业大数据学报, 2020, 2(4): 47-54.
[11]	李冬男, 潘国杨, 周倩, 李斌. 新冠疫情后我国生鲜农产品供应链发展方向[J]. 农业大数据学报, 2020, 2(3): 42-51.
[12]	左敏, 何思宇, 张青川, 姚双顺. 基于区块链的食品溯源技术研究[J]. 农业大数据学报, 2020, 2(3): 52-60.